CN115460484B - User scheduling and resource allocation method and system in federal learning system - Google Patents
User scheduling and resource allocation method and system in federal learning system Download PDFInfo
- Publication number
- CN115460484B CN115460484B CN202211012774.7A CN202211012774A CN115460484B CN 115460484 B CN115460484 B CN 115460484B CN 202211012774 A CN202211012774 A CN 202211012774A CN 115460484 B CN115460484 B CN 115460484B
- Authority
- CN
- China
- Prior art keywords
- optical network
- passive optical
- model
- user
- local
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q11/00—Selecting arrangements for multiplex systems
- H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
- H04Q11/0062—Network aspects
- H04Q11/0067—Provisions for optical access or distribution networks, e.g. Gigabit Ethernet Passive Optical Network (GE-PON), ATM-based Passive Optical Network (A-PON), PON-Ring
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q11/00—Selecting arrangements for multiplex systems
- H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
- H04Q11/0062—Network aspects
- H04Q2011/0086—Network resource allocation, dimensioning or optimisation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Small-Scale Networks (AREA)
- Optical Communication System (AREA)
Abstract
The invention provides a user scheduling and resource allocation method for federal learning traffic in a federal learning system based on a time division wavelength division multiplexing passive optical network, which allocates different transmission wavelengths and sets maximum transmission bandwidths for a local model after user updating according to the marked time point when the aggregation of passive optical bit width units is completed, so as to ensure fair operation, promote the uploading speed of the local model and enable federal learning to complete the training process more quickly.
Description
Technical Field
The invention relates to the technical field of federal learning, in particular to a user scheduling and resource allocation method in a federal learning system based on a time division multiplexing passive optical network and a federal learning system based on the time division multiplexing passive optical network.
Background
In the prior art, the deployment of the 5G mobile network makes the world-wide-interconnection become a reality, and the application of the artificial intelligence technology is also extending from the cloud to the network edge. However, conventional distributed machine learning has failed to meet the architecture of existing networks, and conventional distributed machine learning algorithms typically collect data sets of edge users, and perform training learning in cloud data centers. In general, these data sets are huge, and cause serious collisions to the network, so that the network is congested, and edge users are required to send the data sets to the data center, which inevitably causes data leakage, so that a new distributed machine learning, namely federal learning, is proposed.
Federal learning is a very potential distributed edge learning framework, proposed by Google researchers in 2016, and has received extensive attention from academia and industry. Federal learning allows multiple edge user devices to co-train a model under coordination of a central server, unlike traditional distributed machine learning, edge user devices do not need to send data samples to the central server, only need to update the local model by using local data samples, then upload the trained local model to the central server, the central server aggregates the received local model by using an algorithm, and sends the new local model to the edge user, so that a round of training is completed, and multiple rounds of iterative interactions are performed, so that a model with good accuracy is finally trained.
However, since federal learning requires multiple updating iterations, a large number of users upload their own local training models to the server in each round, which causes a large harm to the network, so that some users are required to select the resource scheduling algorithm, and the harm of federal learning traffic to the network is reduced. The passive optical network system PON architecture can be well compatible with the Federal learning architecture, and the PON system is utilized to realize the Federal learning, so that the passive optical network system PON has the natural advantages which are not possessed by other systems.
Currently, the latest passive optical network is a time division multiplexing passive optical network TWDM-PON, the time division multiplexing passive optical network combines two multiplexes (time division multiplexing and wavelength multiplexing), so as to increase the bandwidth, improve the service quality of users, because the demand for high transmission rate is greater and greater, research on the time division multiplexing passive optical network has attracted attention of more and more expert students, considering the flexibility of wavelengths and access of multiple users, and fairness among units (ONUs) of the network, improving the service quality of users has become necessary, for fairness among ONUs, most students are research based on an ethernet passive optical network, and research on fairness among ONUs is very lacking or even absent in the TWDM-PON network, so reasonable user scheduling and resource allocation are problems to be solved at present in a federal learning system based on the time division multiplexing passive optical network.
Disclosure of Invention
Therefore, the technical problem to be solved by the invention is to overcome the defect that no reasonable user scheduling and resource allocation exist in TWDM-PON networks in the prior art.
In order to solve the technical problems, the invention provides a user scheduling and resource allocation method in a federal learning system based on a time division wavelength division multiplexing passive optical network, which is applied to an optical line terminal and comprises the following steps:
transmitting an initial global model sent by a central server deployed on an optical line terminal to all users deployed on a passive optical network unit;
after the user receives the initial global model and updates the local model, a group of users are randomly selected for marking, and,
the method comprises the steps of carrying out descending order sequencing on wavelengths according to the load condition of the wavelengths, counting the time points of the completion of local model aggregation after user updating marked under each passive optical network unit, carrying out ascending order sequencing on the passive optical network units according to the time points, distributing the wavelengths to each passive optical network unit according to the sequencing result and granting bandwidths;
after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked by each passive optical network unit, so that the central server aggregates the user aggregated model into a new global model, and updating the initial global model;
and judging whether the global loss function is converged or whether the global model updating frequency reaches the upper limit or not, and if not, repeating the steps.
Preferably, when the user receives the initial global model and updates the local model, randomly selecting a group of users for marking includes:
after receiving the initial global model and carrying out local model updating, the users randomly select a group of users and acquire model norms of each user;
and sorting the group of users in a descending order according to the model norm, and selecting a preset number of users with the front sorting order for marking.
Preferably, the calculation formula of the model norm is:wherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->Is the global model of the t-th round.
Preferably, the sorting the wavelengths in descending order according to the loading condition of the wavelengths, counting the time point when the local model aggregation after the user update under each passive optical network unit is completed, sorting the passive optical network units in ascending order according to the time point, and distributing the wavelengths to each passive optical network unit according to the sorting result includes:
ordering wavelengths lambda in descending order according to their loading 1 ,λ 2 ,...,λ n-1 ,λ n ]Wherein n is the total number of wavelengths;
counting the time point [ T ] of the completion of local model aggregation after user updating marked under each passive optical network unit 1 ,T 2 ,T 3 ,...,T k ]Wherein k is the number of passive optical network units;
the passive optical network unit ONU is sequenced in ascending order [ ONU ] according to the time 1 ,ONU 2 ,...,ONU k ]Wherein k is the number of passive optical network units;
allocating wavelength lambda for the passive optical network units according to the sequence numbers of the passive optical network units after sequencing j J=i% n, where i is the passive optical network unit number, j is the wavelength number,
the set of passive optical network units transmitted by each wavelength is W j ＝[ONU j ,ONU n+j ,...,ONU 3n+j ,...]。
Preferably, said granting bandwidth to each passive optical network unit according to the ordering result includes:
calculating the time interval between adjacent passive optical network units on the jth wavelengthWherein x is the passive optical network unit serial number in the passive optical network unit set transmitted on the jth wavelength;
calculating the bandwidth slice size of the passive optical network unit application with serial number x in the passive optical network unit set transmitted on the jth wavelength according to the time intervalAnd-> Wherein Z is the average size of the polling window, FL x Representing the model size, load, of the passive optical network unit transmission with sequence number x j Indicating the load at wavelength j;
and in the passive optical network unit set transmitted on the j-th wavelength, the bandwidth slice size applied by the last passive optical network unit is the maximum allowable bandwidth slice size.
Preferably, the local model updating mode is as followsWherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->For the global model of the t-th round, η is learning rate,>is a local loss function.
Preferably, the aggregation process of the global model is as follows:wherein (1)>Local model updated for the nth round of the nth user, +.>Is the global model of the t+1st round, k is the number of passive optical network units, D u The local data set size for the u-th user, and D is the data set size for all users.
Preferably, the global loss function is:wherein (1)>Represents the global loss function of round t+1,>for the global model of the t-th round, +.>As a local loss function, D u The local data set size of the u-th user, D is the data set size of all users, and k is the number of passive optical network units.
The invention also provides a federal learning system based on the time division multiplexing passive optical network, which comprises:
the passive optical network unit is provided with a plurality of users and is used for aggregating local models uploaded by the users;
the central server is used for initializing training tasks and global models;
the optical line terminal is configured with the central server and is used for transmitting the initial global model sent by the central server to all users deployed on the passive optical network unit, and,
the wavelengths are ordered in descending order according to the load condition of the wavelengths, the time point of the completion of the local model aggregation after the user updating marked under each passive optical network unit is counted, the passive optical network units are ordered in ascending order according to the time point, the wavelengths are distributed to each passive optical network unit and the bandwidth is granted according to the ordering result, and,
after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked under each passive optical network unit so that the central server aggregates the user aggregated model into a new global model, updates the initial global model,
the method is also used for judging whether the current global loss function converges or whether the global model updating frequency reaches the upper limit;
and the optical cable network is used for providing an optical transmission channel between the optical line terminal and the passive optical network unit.
Preferably, the optical line terminal is further configured to randomly select a group of users and obtain a model norm of each user after the users receive the initial global model and update the local model, sort the group of users in descending order according to the model norms, and select a preset number of users with the front sorting for marking.
Compared with the prior art, the technical scheme of the invention has the following advantages:
the invention provides a user scheduling and resource allocation method for federal learning traffic in a federal learning system based on a time division wavelength division multiplexing passive optical network, which allocates different transmission wavelengths and sets maximum transmission bandwidths for a local model after user updating according to the marked time point when the aggregation of passive optical bit width units is completed, so as to ensure fair operation, promote the uploading speed of the local model and enable federal learning to complete the training process more quickly.
Drawings
In order that the invention may be more readily understood, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings, in which:
FIG. 1 is a flow chart of an implementation of a user scheduling and resource allocation method in a Federal learning system based on a time division multiplexing passive optical network of the present invention;
FIG. 2 is a graph comparing model training accuracy results provided by one embodiment;
FIG. 3 is a diagram of resource scheduling and bandwidth allocation provided by one embodiment;
fig. 4 is a federal learning system based on a time division wavelength division multiplexing passive optical network.
Detailed Description
The core of the invention is to provide a user scheduling and resource allocation method in a federal learning system based on a time division multiplexing passive optical network and a federal learning system based on the time division multiplexing passive optical network, which ensure fair operation, promote the uploading speed of a local model and enable federal learning to complete the training process more quickly.
In order to better understand the aspects of the present invention, the present invention will be described in further detail with reference to the accompanying drawings and detailed description. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a user scheduling and resource allocation method in a federal learning system based on a time division wavelength division multiplexing passive optical network according to the present invention; the specific operation steps are as follows:
s101: transmitting an initial global model sent by a central server deployed on an optical line terminal to all users deployed on a passive optical network unit;
s102: after the user receives the initial global model and updates the local model, randomly selecting a group of users for marking;
the local model updating mode is as followsWherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->For the global model of the t-th round, η is learning rate,>is a local loss function.
S103: the method comprises the steps of carrying out descending order sequencing on wavelengths according to the load condition of the wavelengths, counting the time points of the completion of local model aggregation after user updating marked under each passive optical network unit, carrying out ascending order sequencing on the passive optical network units according to the time points, distributing the wavelengths to each passive optical network unit according to the sequencing result and granting bandwidths;
ordering wavelengths lambda in descending order according to their loading 1 ,λ 2 ,...,λ n-1 ,λ n ]Wherein n is the total number of wavelengths;
counting the time point [ T ] of the completion of local model aggregation after user updating marked under each passive optical network unit 1 ,T 2 ,T 3 ,...,T k ]Wherein k is the number of passive optical network units;
the passive optical network unit ONU is sequenced in ascending order [ ONU ] according to the time 1 ,ONU 2 ,...,ONU k ]Wherein k is the number of passive optical network units;
allocating wavelength lambda for the passive optical network units according to the sequence numbers of the passive optical network units after sequencing j J=i% n, where i is the passive optical network unit number, j is the wavelength number,
the set of passive optical network units transmitted by each wavelength is W j ＝[ONU j ,ONU n+j ,...,ONU 3n+j ,...]。
Calculating the time interval between adjacent passive optical network units on the jth wavelengthWherein x is the passive optical network unit serial number in the passive optical network unit set transmitted on the jth wavelength;
calculating the bandwidth slice size of the passive optical network unit application with serial number x in the passive optical network unit set transmitted on the jth wavelength according to the time intervalAnd to ensure that the bandwidth slot granted to that ONU is too large to interfere with other ONUs: /> Determined by the wavelength loading: />Wherein Z is the average size of the polling window, FL x Representing the model size, load, of the passive optical network unit transmission with sequence number x j Indicating the load at wavelength j;
and in the passive optical network unit set transmitted on the j-th wavelength, the bandwidth slice size applied by the last passive optical network unit is the maximum allowable bandwidth slice size.
S104: after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked by each passive optical network unit, so that the central server aggregates the user aggregated model into a new global model, and updating the initial global model;
the aggregation process of the global model is as follows:wherein (1)>Local model updated for the nth round of the nth user, +.>Is the global model of the t+1st round, k is the number of passive optical network units, D u The local data set size for the u-th user, and D is the data set size for all users.
S105: and judging whether the global loss function is converged or whether the global model updating frequency reaches the upper limit or not, and if not, repeating the steps.
The global loss function is:wherein (1)>Represents the global loss function of round t+1,>global model for the t-th roundIs (1)>As a local loss function, D u The local data set size of the u-th user, D is the data set size of all users, and k is the number of passive optical network units.
The invention provides a user scheduling and resource allocation method for federal learning traffic in a federal learning system based on a time division wavelength division multiplexing passive optical network, which allocates different transmission wavelengths for a local model after user updating according to the time point when passive optical bit width unit aggregation is completed, and sets the maximum transmission bandwidth according to the load of the wavelength so as to ensure fair operation, thereby improving the uploading speed of the local model and enabling federal learning to complete the training process more quickly.
Based on the above embodiments, the present embodiment further describes step S102:
after the user receives the initial global model and updates the local model, randomly selecting a group of users and acquiring a model norm of each user, wherein the calculation formula of the model norms is as follows:wherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->Is the global model of the t-th round;
and sorting the group of users in a descending order according to the model norm, and selecting a preset number of users with the front sorting order for marking.
The invention carries out the second selection based on random selection, and uploads the user model with larger norm value, thereby improving the convergence rate of the model.
Based on the above embodiment, the verification is performed on an open source platform according to the embodiment, which is specifically as follows:
the method comprises the steps of adopting 16 passive Optical Network Units (ONU), wherein the same number of users exist under each ONU, the number of wavelengths is 4, the wavelength load normalization values are 0.1,0.3,0.5 and 0.7, the number of ONU under each wavelength is consistent, the distance between an Optical Line Terminal (OLT) and the ONU is set to 20Km, the downloading rate is set to 10Gb/s, and the uploading single wavelength rate is set to 2.5Gb/s.
Training with a two layer 5*5 CNN network contains 26.4 megabytes of parameters, assuming each parameter is quantized to 1bit information, and thus the amount of data per model is 26.4 megabytes. The learning rate is set to 0.006, and the mi i-batch size for user local model updates is set to 10, with local updates being performed once per round. As shown in fig. 2, 120 rounds of federal learning are performed, and the accuracy result of the random selection strategy is compared with that of the user with the number of 32, it can be seen that, compared with the random user selection strategy, the user selection strategy convergence speed is faster, the accuracy is higher, different numbers of clients participate in training, the accuracy is obviously affected, and meanwhile, the uploading time is shortened.
As shown in fig. 3, which is a schematic diagram of resource scheduling and bandwidth allocation, the four wavelength loading conditions are respectively 0.7, 0.5, 0.3, and 0.1, and the ONU1 to ONU16 are time-ordered ONU serial numbers corresponding to the wavelengths w1 to w4, it can be seen that if the ONU with long training time transmits on the wavelength with heavy loading, the training time will be increased, where the ONU length in the figure indicates the allocated bandwidth.
As shown in fig. 4, the present invention further provides a federal learning system based on a time division wavelength division multiplexing passive optical network, including:
the passive optical network unit ONU is deployed with a plurality of users and is used for aggregating local models uploaded by the users;
the central server is used for initializing training tasks and global models;
the optical line terminal OLT is configured with the central server and is used for transmitting the initial global model sent by the central server to all users deployed on the passive optical network unit, and,
the wavelengths are ordered in descending order according to the load condition of the wavelengths, the time point of the completion of the local model aggregation after the user updating marked under each passive optical network unit is counted, the passive optical network units are ordered in ascending order according to the time point, the wavelengths are distributed to each passive optical network unit and the bandwidth is granted according to the ordering result, and,
after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked under each passive optical network unit so that the central server aggregates the user aggregated model into a new global model, updates the initial global model,
the method is also used for judging whether the current global loss function converges or whether the global model updating frequency reaches the upper limit;
and the optical cable network ODN is used for providing an optical transmission channel between the optical line terminal and the passive optical network unit.
It is apparent that the above examples are given by way of illustration only and are not limiting of the embodiments. Other variations and modifications of the present invention will be apparent to those of ordinary skill in the art in light of the foregoing description. It is not necessary here nor is it exhaustive of all embodiments. While still being apparent from variations or modifications that may be made by those skilled in the art are within the scope of the invention.
Claims (7)
1. A method for user scheduling and resource allocation in a federal learning system, applied to an optical line terminal, comprising:
transmitting an initial global model sent by a central server deployed on an optical line terminal to all users deployed on a passive optical network unit;
after the user receives the initial global model and updates the local model, a group of users are randomly selected for marking, and,
the method comprises the steps of carrying out descending order sequencing on wavelengths according to the load condition of the wavelengths, counting the time points of the completion of local model aggregation after user updating marked under each passive optical network unit, carrying out ascending order sequencing on the passive optical network units according to the time points, distributing the wavelengths to each passive optical network unit according to the sequencing result and granting bandwidths;
after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked by each passive optical network unit, so that the central server aggregates the user aggregated model into a new global model, and updating the initial global model;
judging whether the global loss function is converged or whether the global model updating frequency reaches the upper limit or not, if not, repeating the steps;
the local model updating mode is as followsWherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->For the global model of the t-th round, η is learning rate,>is a local loss function;
the aggregation process of the global model is as follows:wherein (1)>Local model updated for the nth round of the nth user, +.>Is the global model of the t+1st round, k is the number of passive optical network units, D u The local data set size of the u-th user, and the data set size of all users;
the saidThe calculation formula of the global loss function is as follows:wherein (1)>Represents the global loss function of round t+1,>for the global model of the t-th round, +.>As a local loss function, D u The local data set size of the u-th user, D is the data set size of all users, and k is the number of passive optical network units.
2. The method of claim 1, wherein randomly selecting a group of users for tagging after the users receive the initial global model and perform local model updates comprises:
after receiving the initial global model and carrying out local model updating, the users randomly select a group of users and acquire model norms of each user;
and sorting the group of users in a descending order according to the model norm, and selecting a preset number of users with the front sorting order for marking.
3. A method of user scheduling and resource allocation in a federal learning system according to claim 2, wherein the model norm is calculated as:wherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->Is the global model of the t-th round.
4. The method for scheduling and allocating resources in a federal learning system according to claim 1, wherein the step of ordering the wavelengths in descending order according to the loading condition of the wavelengths, counting the time point when the local model aggregation after the user update marked under each passive optical network unit is completed, ordering the passive optical network units in ascending order according to the time point, and allocating the wavelengths to each passive optical network unit according to the ordering result comprises:
ordering wavelengths lambda in descending order according to their loading 1 ,λ 2 ,,λ n-1 ,λ n ]Wherein n is the total number of wavelengths;
counting the time point [ T ] of the completion of local model aggregation after user updating marked under each passive optical network unit 1 ,T 2 ,T 3 ,,T k ]Wherein k is the number of passive optical network units;
the passive optical network unit ONU is sequenced in ascending order [ ONU ] according to the time 1 ,ONU 2 ,...,ONU k ]Wherein k is the number of passive optical network units;
allocating wavelength lambda for the passive optical network units according to the sequence numbers of the passive optical network units after sequencing j J=i% n, where i is the passive optical network unit number, j is the wavelength number,
the set of passive optical network units transmitted by each wavelength is W j ＝[ONU j ,ONU n+j ,...,ONU 3n+j ,...]。
5. The method for user scheduling and resource allocation in a federal learning system according to claim 4, wherein said granting bandwidth to each of said passive optical network units according to the ordering result comprises:
calculating adjacent passives at the jth wavelengthTime interval between optical network unitsWherein x is the passive optical network unit serial number in the passive optical network unit set transmitted on the jth wavelength;
calculating the bandwidth slice size of the passive optical network unit application with serial number x in the passive optical network unit set transmitted on the jth wavelength according to the time intervalAnd->Wherein Z is the average size of the polling window, FL x Representing the model size, load, of the passive optical network unit transmission with sequence number x j Indicating the load at wavelength j;
and in the passive optical network unit set transmitted on the j-th wavelength, the bandwidth slice size applied by the last passive optical network unit is the maximum allowable bandwidth slice size.
6. A federal learning system based on a time division multiplexing passive optical network, comprising:
the passive optical network unit is provided with a plurality of users and is used for aggregating local models uploaded by the users;
the central server is used for initializing training tasks and global models;
the optical line terminal is configured with the central server and is used for transmitting the initial global model sent by the central server to all users deployed on the passive optical network unit, and,
the wavelengths are ordered in descending order according to the load condition of the wavelengths, the time point of the completion of the local model aggregation after the user updating marked under each passive optical network unit is counted, the passive optical network units are ordered in ascending order according to the time point, the wavelengths are distributed to each passive optical network unit and the bandwidth is granted according to the ordering result, and,
after the wavelength and bandwidth allocation is finished, acquiring a user aggregated model marked under each passive optical network unit so that the central server aggregates the user aggregated model into a new global model, updates the initial global model,
the method is also used for judging whether the current global loss function converges or whether the global model updating frequency reaches the upper limit;
an optical cable network for providing an optical transmission channel between the optical line terminal and the passive optical network unit;
the local model updating mode is as followsWherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->For the global model of the t-th round, η is learning rate,>is a local loss function;
the aggregation process of the global model is as follows:wherein (1)>Local model updated for the nth round of the nth user, +.>Is the global model of the t+1st round, k is the number of passive optical network units, D u Local data set size for the u-th user, D isThe data set size of all users; the calculation formula of the global loss function is as follows:wherein (1)>Represents the global loss function of round t+1,>as a global model of the t-th round,as a local loss function, D u The local data set size of the u-th user, D is the data set size of all users, and k is the number of passive optical network units.
7. The federal learning system based on a time division wavelength division multiplexing passive optical network according to claim 6, wherein the optical line terminal is further configured to, when a user receives the initial global model and performs local model update, randomly select a group of users and obtain a model norm of each user therein, sort the group of users in descending order according to the model norms, and select a preset number of users ranked in front for marking; the calculation formula of the model norm is as follows:wherein (1)>Local model updated for the (t+1) -th round of the (u) -th user,/->Is the global model of the t-th round.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211012774.7A CN115460484B (en) | 2022-08-23 | 2022-08-23 | User scheduling and resource allocation method and system in federal learning system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211012774.7A CN115460484B (en) | 2022-08-23 | 2022-08-23 | User scheduling and resource allocation method and system in federal learning system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115460484A CN115460484A (en) | 2022-12-09 |
| CN115460484B true CN115460484B (en) | 2023-07-21 |
Family
ID=84298431
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211012774.7A Active CN115460484B (en) | 2022-08-23 | 2022-08-23 | User scheduling and resource allocation method and system in federal learning system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115460484B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111580970A (en) * | 2020-05-07 | 2020-08-25 | 电子科技大学 | Transmission scheduling method for model distribution and aggregation of federated learning |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210365841A1 (en) * | 2020-05-22 | 2021-11-25 | Kiarash SHALOUDEGI | Methods and apparatuses for federated learning |
| CN113660112B (en) * | 2021-07-23 | 2023-11-24 | 广州大学 | Federal learning bandwidth allocation method, system, computer equipment and medium |
| CN114554495A (en) * | 2022-01-29 | 2022-05-27 | 南京邮电大学 | Federal learning-oriented user scheduling and resource allocation method |
-
2022
- 2022-08-23 CN CN202211012774.7A patent/CN115460484B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111580970A (en) * | 2020-05-07 | 2020-08-25 | 电子科技大学 | Transmission scheduling method for model distribution and aggregation of federated learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115460484A (en) | 2022-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111277437B (en) | Network slice resource allocation method for smart power grid | |
| CN108684046B (en) | Random learning-based access network service function chain deployment method | |
| KR102398021B1 (en) | Method and apparatus for allocating bandwidth based on machine learning in passive optical network | |
| CN108834004B (en) | Routing calculation, fiber core selection and frequency spectrum allocation method and system based on cross crosstalk perception | |
| CN103188279B (en) | By the method and apparatus that peer-to-peer network downloads file from multiple neighbor nodes | |
| CN110996196B (en) | Optimal route optimizing method for optical transmission network optical path fiber core utilization | |
| CN105119682A (en) | User level protocol and QoS (Quality of Service) based dynamic wavelength bandwidth distribution method in WDM-EPON (Wavelength Division Multiplexing-Ethernet Passive Optical Network) | |
| CN109743751B (en) | Resource allocation method and device for wireless access network | |
| CN101409596A (en) | Method for planning dynamic business wavelength route optical network | |
| CN115103450B (en) | Multi-service time slot allocation method and equipment | |
| CN114268548A (en) | Network slice resource arranging and mapping method based on 5G | |
| US8089985B2 (en) | Max-Min fair network bandwidth allocator | |
| Li et al. | Scalable federated learning over passive optical networks | |
| CN113094180B (en) | Wireless federal learning scheduling optimization method and device | |
| CN115460484B (en) | User scheduling and resource allocation method and system in federal learning system | |
| CN113015040B (en) | Resource allocation method based on fragment and domain matching degree in multi-core elastic optical network | |
| CN113225824A (en) | Device and method for automatically allocating bandwidths with different service requirements based on 5G technology | |
| Ciceri et al. | Federated learning over next-generation ethernet passive optical networks | |
| CN116074260A (en) | Service slice scheduling method in power network | |
| CN113867736B (en) | Deployment scheme generation method and device | |
| CN115329987B (en) | User selection method in federal learning system | |
| CN114585093A (en) | Method, device and equipment for sending service data | |
| CN114390489A (en) | Service deployment method for end-to-end network slice | |
| Li et al. | Bandwidth slicing to boost federated learning in edge computing | |
| CN104346655B (en) | Application method of honeycomb structure dynamic algorithm to intelligent power utilization information management system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |